]> granicus.if.org Git - postgresql/commitdiff
Improve test coverage of pg_rewind
authorMichael Paquier <michael@paquier.xyz>
Tue, 8 Oct 2019 02:46:30 +0000 (11:46 +0900)
committerMichael Paquier <michael@paquier.xyz>
Tue, 8 Oct 2019 02:46:30 +0000 (11:46 +0900)
This includes new TAP tests for a couple of areas not covered yet and
some improvements:
- More coverage for --no-ensure-shutdown, the enforced recovery step and
--dry-run.
- Failures with option combinations and basic option checks.
- Removal of a duplicated comment.

Author: Alexey Kondratov, Michael Paquier
Discussion: https://postgr.es/m/20191007010651.GD14532@paquier.xyz

src/bin/pg_rewind/t/001_basic.pl
src/bin/pg_rewind/t/005_same_timeline.pl
src/bin/pg_rewind/t/006_options.pl [new file with mode: 0644]

index c3293e93df7c32839162683eef1d04785199fd71..95d8ccfced1b2fe3b2151159eaa9f8b80d66929e 100644 (file)
@@ -1,7 +1,7 @@
 use strict;
 use warnings;
 use TestLib;
-use Test::More tests => 11;
+use Test::More tests => 15;
 
 use FindBin;
 use lib $FindBin::RealBin;
@@ -66,6 +66,71 @@ sub run_test
        master_psql("DELETE FROM tail_tbl WHERE id > 10");
        master_psql("VACUUM tail_tbl");
 
+       # Before running pg_rewind, do a couple of extra tests with several
+       # option combinations.  As the code paths taken by those tests
+       # do not change for the "local" and "remote" modes, just run them
+       # in "local" mode for simplicity's sake.
+       if ($test_mode eq 'local')
+       {
+               my $master_pgdata  = $node_master->data_dir;
+               my $standby_pgdata = $node_standby->data_dir;
+
+               # First check that pg_rewind fails if the target cluster is
+               # not stopped as it fails to start up for the forced recovery
+               # step.
+               command_fails(
+                       [
+                               'pg_rewind',       '--debug',
+                               '--source-pgdata', $standby_pgdata,
+                               '--target-pgdata', $master_pgdata,
+                               '--no-sync'
+                       ],
+                       'pg_rewind with running target');
+
+               # Again with --no-ensure-shutdown, which should equally fail.
+               # This time pg_rewind complains without attempting to perform
+               # recovery once.
+               command_fails(
+                       [
+                               'pg_rewind',       '--debug',
+                               '--source-pgdata', $standby_pgdata,
+                               '--target-pgdata', $master_pgdata,
+                               '--no-sync',       '--no-ensure-shutdown'
+                       ],
+                       'pg_rewind --no-ensure-shutdown with running target');
+
+               # Stop the target, and attempt to run with a local source
+               # still running.  This fails as pg_rewind requires to have
+               # a source cleanly stopped.
+               $node_master->stop;
+               command_fails(
+                       [
+                               'pg_rewind',       '--debug',
+                               '--source-pgdata', $standby_pgdata,
+                               '--target-pgdata', $master_pgdata,
+                               '--no-sync',       '--no-ensure-shutdown'
+                       ],
+                       'pg_rewind with unexpected running source');
+
+               # Stop the target cluster cleanly, and run again pg_rewind
+               # with --dry-run mode.  If anything gets generated in the data
+               # folder, the follow-up run of pg_rewind will most likely fail,
+               # so keep this test as the last one of this subset.
+               $node_standby->stop;
+               command_ok(
+                       [
+                               'pg_rewind',       '--debug',
+                               '--source-pgdata', $standby_pgdata,
+                               '--target-pgdata', $master_pgdata,
+                               '--no-sync',       '--dry-run'
+                       ],
+                       'pg_rewind --dry-run');
+
+               # Both clusters need to be alive moving forward.
+               $node_standby->start;
+               $node_master->start;
+       }
+
        RewindTest::run_pg_rewind($test_mode);
 
        check_query(
index df469d393939634a62070346936393223d47c6b8..5464f4203a70ff33c18a219117d1db8b2f28c606 100644 (file)
@@ -12,9 +12,6 @@ use lib $FindBin::RealBin;
 
 use RewindTest;
 
-# Test that running pg_rewind if the two clusters are on the same
-# timeline runs successfully.
-
 RewindTest::setup_cluster();
 RewindTest::start_master();
 RewindTest::create_standby();
diff --git a/src/bin/pg_rewind/t/006_options.pl b/src/bin/pg_rewind/t/006_options.pl
new file mode 100644 (file)
index 0000000..1515696
--- /dev/null
@@ -0,0 +1,40 @@
+#
+# Test checking options of pg_rewind.
+#
+use strict;
+use warnings;
+use TestLib;
+use Test::More tests => 12;
+
+program_help_ok('pg_rewind');
+program_version_ok('pg_rewind');
+program_options_handling_ok('pg_rewind');
+
+my $primary_pgdata = TestLib::tempdir;
+my $standby_pgdata = TestLib::tempdir;
+command_fails(
+       [
+               'pg_rewind',       '--debug',
+               '--target-pgdata', $primary_pgdata,
+               '--source-pgdata', $standby_pgdata,
+               'extra_arg1'
+       ],
+       'too many arguments');
+command_fails([ 'pg_rewind', '--target-pgdata', $primary_pgdata ],
+       'no source specified');
+command_fails(
+       [
+               'pg_rewind',       '--debug',
+               '--target-pgdata', $primary_pgdata,
+               '--source-pgdata', $standby_pgdata,
+               '--source-server', 'incorrect_source'
+       ],
+       'both remote and local sources specified');
+command_fails(
+       [
+               'pg_rewind',       '--debug',
+               '--target-pgdata', $primary_pgdata,
+               '--source-pgdata', $standby_pgdata,
+               '--write-recovery-conf'
+       ],
+       'no local source with --write-recovery-conf');